CodSpeed Performance Report

Merging #2373 will not alter performance

_{Comparing gjcarneiro:feature/parent-scoped-naming (46a7da6) with main (b0d9109)}

Summary

✅ 31 untouched benchmarks

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.18%. Comparing base (117a31d) to head (0654506).

@@           Coverage Diff           @@
##             main    #2373   +/-   ##
=======================================
  Coverage   98.17%   98.18%           
=======================================
  Files          65       65           
  Lines        7945     7968   +23     
  Branches      800      804    +4     
=======================================
+ Hits         7800     7823   +23     
  Misses        110      110           
  Partials       35       35           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

@koxudaxi This is a very useful flag to have. Could this be merged? 🙌

I think I am having the same issue as @gjcarneiro with a model having many Spec or Status fields defined in an openapi spec, which is super common in kubernetes-based resources

Example: spec, swagger ui

But unfortunately, using the --parent-scoped-naming in my case led to a very nested file structure of the generated model with some models even failing to build hitting the file name length limits:

            cmd = [
                "datamodel-codegen",
                "--url",
                spec["url"],
                "--output-model-type",
                "pydantic_v2.BaseModel",
                "--use-annotated",
                "--parent-scoped-naming",
                "--enum-field-as-literal",
                "all",
                "--output",
                str(output_dir),
            ]

I wonder if I hold it wrong? I thought this would result on class names to be prepended with the ancestor info, but not lead to the file structure changes.

Here is what I meant by saying that the file tree gets exploded. When I generate the model using this new flag I get this:

as opposed to a much flatter structure before

I'm sorry if this option does not work correctly for your case. I did not test this with any api specs that contain dots. In my use cases, I always only have a single flat python module, without any dirs.

You can check around this area what it is trying to do. It simply concatenates the path of parent of the current model to the current model's path. Perhaps it needs to be slightly tweaked when there is a "." (dot) in the path.